@ngzhian, I believe the simd_i64x2_cmp.wast tests could be improved if we varied the expected results of the ge_s and le_s tests. Most of those tests expect [-1, -1] so when I made a mistake re: flipped operands the tests did not fail for me. When I added tests like the following then I observed the expected failure (note the expected result is not all 1s):
Should we add a few of these types of tests (for implementor sanity)? And have you been adding these by hand or using some test generator (I think there was a Python one somewhere)?
@ngzhian, I believe the simd_i64x2_cmp.wast tests could be improved if we varied the expected results of the
ge_s
andle_s
tests. Most of those tests expect[-1, -1]
so when I made a mistake re: flipped operands the tests did not fail for me. When I added tests like the following then I observed the expected failure (note the expected result is not all 1s):Should we add a few of these types of tests (for implementor sanity)? And have you been adding these by hand or using some test generator (I think there was a Python one somewhere)?